Re: qemu-4.0.1: vhost_region_add_section:Section rounded to 0 prior to previous a0000

[Peter Lieven]

> Am 16.01.2020 um 21:26 schrieb Dr. David Alan Gilbert <address@hidden>:
> 
> * Peter Lieven (address@hidden) wrote:
>> Am 16.01.20 um 13:47 schrieb Peter Lieven:
>>> Am 13.01.20 um 17:25 schrieb Peter Lieven:
>>>> Am 09.01.20 um 19:44 schrieb Dr. David Alan Gilbert:
>>>>> * Peter Lieven (address@hidden) wrote:
>>>>>> Am 08.01.20 um 16:04 schrieb Dr. David Alan Gilbert:
>>>>>>> * Peter Lieven (address@hidden) wrote:
>>>>>>>> Hi,
>>>>>>>> 
>>>>>>>> 
>>>>>>>> I have a Qemu 4.0.1 machine with vhost-net network adapter, thats 
>>>>>>>> polluting the log with the above message.
>>>>>>>> 
>>>>>>>> Is this something known? Googling revealed the following patch in Nemu 
>>>>>>>> (with seems to be a Qemu fork from Intel):
>>>>>>>> 
>>>>>>>> https://github.com/intel/nemu/commit/03940ded7f5370ce7492c619dccced114ef7f56e
>>>>>>>> 
>>>>>>>> 
>>>>>>>> The network stopped functioning. After a live-migration the vServer is 
>>>>>>>> reachable again.
>>>>>>>> 
>>>>>>>> 
>>>>>>>> Any ideas?
>>>>>>> What guest are you running and what does your qemu commandline look
>>>>>>> like?
>>>>>> 
>>>>>> Its running debian9. We have hundreds of other VMs with identical setup. 
>>>>>> Do not know why this one makes trouble.
>>>>> Could you extract an 'info mtree' from it - particularly the
>>>>> 'address-space: memory' near the top.
>>>> 
>>>> 
>>>> Here we go:
>>>> 
>>>> 
>>>> address-space: memory
>>>>   0000000000000000-ffffffffffffffff (prio 0, i/o): system
>>>>     0000000000000000-000000003fffffff (prio 0, i/o): alias ram-below-4g 
>>>> @pc.ram 0000000000000000-000000003fffffff
>>>>     0000000000000000-ffffffffffffffff (prio -1, i/o): pci
>>>>       00000000000a0000-00000000000affff (prio 2, i/o): alias vga.chain4 
>>>> @vga.vram 0000000000000000-000000000000ffff
>>>>       00000000000a0000-00000000000bffff (prio 1, i/o): vga-lowmem
>>> 
>>> 
>>> What seems special is that the RAM area is prio2. Any idea if this makes 
>>> trouble?
>> 
>> 
>> Update from my side. This happens when I have Debian 10 with XFCE when the 
>> Graphical User Interface is initialized.
>> 
>> I see the log message when I specify -M pc-i440fx-2.9. If I obmit the 
>> machine type the error does not appear.
> 
> I can't persuade this to reproduce here on the images I currently have;
> but if you can rebuild, can you try the v3 of 'Fix hyperv synic on
> vhost' I've just posted?  It turns off the alignment code that's
> spitting that error in vhost-kernel cases, so should go away.


I will definitely give it a try tomorrow. The fix to change the machine type 
was not working. I was too fast with my assumption.

What I see is the following:

Machine boots up (cold start), has network connectivity. As soon as the 
Graphics are initialized, the vhost_region_add_section error pops up and
Networking is interrupted. When I then migrate the vServer to another host 
Networking works again. It even seems to keep working when I do a
soft reset (shutdown -r). The only thing that breaks networking again when the 
graphic is initialized is a hard reset (system reset via hmp or amp).

Thank you,
Peter